Many of the best AI Agent plays are in areas where you aim to automate the full workflow, but there’s plenty of value in automating individual tasks or chunks of work along the way. You can create a ton of value now, but then the TAM just grows as the models get better.

One practice we’ve implemented in going AI-first at Box is every week someone demos their AI workflow to the whole company. Usually it’s something we never would’ve predicted. Good to balance central planning for tech and decentralization for use case adoption.

NVIDIA will be the first $10 trillion dollar company

The best AI Agent plays are where the model performs well, but where there’s still a big gap between the model and the workflow. This is where you can build a real moat with better domain understanding, data, tool use, context, and a tailored UX. Many opportunities here.

There’s likely too much fear that AI models eat the app layer as they improve. For AI Agents to work, most enterprises will require a bridge between the AI and their specific workflows. It turns out the last mile of making AI Agents work in real, highly variable and hostile environments, is insanely hard. And increasingly it’s the most valuable part of the whole process. That bridge between AI models and enterprise workflows will be a heavy amount of software to connect to different systems, pulling in the right enterprise data, handling security and permissions properly, and having a deep level of context tied to the use case. Then you add in customer support tailored to the use-case, SLAs, liability clauses, tailored sales motions, aligned partnerships for the category, and so on. The list required is quite endless. Every single vertical, and even critical horizontal category, will require a deep amount of expertise to make the AI Agents effective. The big opportunity right now is to identify where these gaps are the widest (between model and the workflow), and fill them in with the appropriate software and expertise. And, even as the models improve - which has previously presented the risk of cannibalization - the focused players can just offer even more value and use-cases to customers. There’s almost no scenario, if you’ve gone after the right market opportunity, where model improvements are a bad thing when building AI Agents.

There is so much opportunity right now in making AI Agents real for enterprises. The models pack insane capabilities, but to get them into the middle of an enterprise workflow requires significant scaffolding, domain understanding, and context engineering. This is the play.

Building AI Agents is an exercise in balancing deterministic and non-deterministic workflows. Too little agent free rein and you don’t benefit from the model’s intelligence. Too much and you’ll get security issues or haphazard results. This is why SaaS + Agents will remain.

If you’re building AI Agents you have to find the optimal level of work to automate for max effectiveness and TAM size. Too narrow and the agents will work but the TAM is small. Too general and the agents won’t work yet but the TAM is big. Have to find the sweet spot.

Context engineering is increasingly the most critical component for building effective AI Agents in the enterprise right now. This will ultimately be the long pole in the tent for AI Agents adoption in most organizations. We need AI Agents that can deeply understand the context of the business process that they’re tied to. This means accessing the most important data for that workflow, using the appropriate tools at the right moment, having proper objectives and instructions, and understanding the domain that they’re in. Some of the big open items for anyone building enterprise agents are: * Narrow vs. General agents. The smaller the task, the easier it is to give the AI Agents the right context to be successful. But the smaller the task, the less value there will be. Finding the optimal task size for value generation will be an important factor for the next few years. * Getting data into an agent-ready system. Enterprise data is often fragmented between dozens or hundreds of systems, many of which are not prepared for a world of AI. Most companies will still need to modernize their data environments to get the full benefit of AI Agents. * Accessing the *right* data for the task is paramount. Even when you have data in a modern environment, getting access controls perfectly aligned to what the AI Agent is going to need access to is critical. Further, deciding what to do RAG on vs. just a general search vs. what to put fully into the context window will matter a ton per task. * Choosing what should be deterministic vs. non-deterministic. If you demand too much from the models, you’re likely to see some drop off in quality. Yet, if you have the model do too little, then you’re dramatically underutilizing what’s possible with AI. This of course is a moving target because the models themselves are improving at an accelerating rate. * The right user interface to get the AI Agents context deeply matters. Half of the problem for getting context to agents doesn’t look like an AI problem at all. It’s all about where the agents show up in the workflow and how the user interacts with them to provide them the context necessary to do the task. The race for the next few years in AI in the enterprise is to see who best to deliver the right context for any given workflow. This will determine the winners and losers in the AI race.

It’s wild how many AI Agent opportunities there are that are much bigger than their pre-AI markets. The TAM of the code editor market was a few billion dollars in 2020. AI coding agents could be a $50-100B+ market in the next 5-10 years. You can do this for many spaces.